With the increasing security demands in public spaces, rapid and precise weapon identification has become crucial for modern surveillance systems. This paper presents a two-stage pipeline model for real-time weapon detection in CCTV footage, aiming to enhance public safety in environments like events, and crowded areas. The proposed approach combines YOLOv11 and Faster R- CNN to balance speed and precision in weapon detection. In the first stage, YOLOv11 functions as a preliminary detector, quickly scanning each frame to identify potential regions of interest (ROIs) where weapons might be present. The bounding boxes and confidence scores of these ROIs are then passed to the second stage. Here, Faster R-CNN refines the detection, analyzing the identified ROIs with greater accuracy, which is especially beneficial for smaller or more challenging-to-detect objects. The results from two-stage pipeline model effectively combines YOLOv11’s speed with Faster R-CNN’s detailed accuracy, demonstrating a robust solution for real- time, automated threat recognition in surveillance settings.